A 3D transistor architecture that promises higher AI computing performance, lower power consumption and significantly denser chip designs for future electronics.
IBM has unveiled what it calls the world’s first semiconductor technology capable of producing chips smaller than 1 nanometer, marking a major milestone in advanced electronics and AI hardware. The new 0.7-nanometer (7-angstrom) chip technology is designed to deliver substantially higher transistor density while improving computing performance and energy efficiency, paving the way for next-generation processors used in artificial intelligence, cloud infrastructure and edge devices.
The breakthrough is based on IBM’s newly developed Nanostack architecture, which departs from conventional transistor layouts by stacking transistors vertically instead of arranging them side by side. This three-dimensional design enables nearly 100 billion transistors to fit onto a chip roughly the size of a fingernail—almost doubling the transistor density achieved by IBM’s 2-nanometer technology introduced in 2021. According to the company, the architecture can deliver up to 50% higher performance or reduce power consumption by as much as 70% compared with its previous 2nm node.
Beyond transistor scaling, IBM has also redesigned on-chip static random-access memory (SRAM), reducing its footprint by approximately 40%. Since SRAM occupies a significant portion of modern AI processors, the improvement allows designers to integrate more cache memory or additional computing cores within the same silicon area. This enhancement is expected to improve data movement and processing efficiency in AI accelerators that support large language models, generative AI workloads and high-performance computing applications.
The development comes as semiconductor manufacturers seek alternatives to traditional chip scaling to sustain Moore’s Law amid growing physical limitations. IBM’s vertical transistor architecture demonstrates that performance gains remain achievable through structural innovation rather than simple dimensional shrinking. The technology also strengthens IBM’s position in semiconductor research alongside industry leaders developing advanced process nodes for future AI systems.
Although the technology remains at the research stage, IBM expects commercial production could begin within the next five years through manufacturing partners. The company has previously licensed advanced chip technologies to Samsung and Japan’s Rapidus, but it has not yet announced a fabrication partner for the new sub-1nm platform. If successfully commercialized, the innovation could enable faster, more energy-efficient processors for data centers, consumer electronics and future AI-driven computing systems.
